The 'classic' Perlish approach to this type of problem involves bitwise string boolean operations. The string  $diff generated by the bitwise-xor of characters in original sequence strings can be used to produce masks that can then be used to extract the differing sub-string sequences from the original strings.

use warnings;
use strict;

my $s1 = 'ACTGGACGTATGCA';
my $s2 = 'AGTG-ACGC-CGCA';

my $diff = $s1 ^ $s2;

my @dpos;
push @dpos, [ $-[1], $+[1] - $-[1] ]
    while $diff =~ m{ ([^\x00]+) }xmsg;

print qq{diff at offset $_->[0], length $_->[1] \n}
    for @dpos;

(my $mask = $diff) =~ tr{\x00}{\xff}c;

$s1 &= $mask;
$s2 &= $mask;

my $differences = qr{ [^\x00]+ }xms;

@dpos = ();
while ($s1 =~ m{ ($differences) }xmsg) {
    # this code produces same result
    # my @diff_data = ($1);
    # $s2 =~ m{ ($differences) }xmsg;
    # push @diff_data, $1, $-[1];
    # push @dpos, \@diff_data;
    push @dpos,
         [ $1, do { $s2 =~ m{ ($differences) }xmsg && $1, $-[1] } ]
         ;
    }

print qq{@$_ \n} for @dpos;
[download]

Output:

diff at offset 1, length 1
diff at offset 4, length 1
diff at offset 8, length 3
C G 1
G - 4
TAT C-C 8
[download]

See @- and @+ in perlvar, also Bitwise Or and Exclusive Or and Bitwise And in perlop.

BrowserUk is very good on this general topic.

Update: Added better code example, doc links. And thanks to ELISHEVA.

Update: Fixed @- link above. What was I thinking?